Wield: Systematic Reinforcement Learning With Progressive Randomization
This addresses the challenge for practitioners in reinforcement learning by providing a systematic approach to task design, though it appears incremental as it builds on existing frameworks.
The paper tackles the problem of task design in reinforcement learning by introducing Wield, a system that decouples configuration from state and action design, resulting in a novel protocol using staged randomization for incremental evaluation.
Reinforcement learning frameworks have introduced abstractions to implement and execute algorithms at scale. They assume standardized simulator interfaces but are not concerned with identifying suitable task representations. We present Wield, a first-of-its kind system to facilitate task design for practical reinforcement learning. Through software primitives, Wield enables practitioners to decouple system-interface and deployment-specific configuration from state and action design. To guide experimentation, Wield further introduces a novel task design protocol and classification scheme centred around staged randomization to incrementally evaluate model capabilities.